We host a #Minimax #MinimaxH3 demo page running #Veda Sparse Attention on our own GPUs. Welcome to try it out! Feel free to share your results~
> veda.hanxiao.run <
ComfyUI Plugin:github.com/veda-sparse/Veda-…
Adaptor:huggingface.co/Veda-Sparse/M…
#Veda Sparse Attention ComfyUI node for #MiniMaxH3 is coming soon :) Up to 3x end-to-end speedup with quality comparable to full attention, on a single RTX 4090.
huggingface.co/Veda-Sparse/M…
It works across all modalities (T2VA/FL2VA/REF2VA), with any LoRA and any step count.
HanxiaoMeow retweeted
A quick test of the @M5Stack PaperMono. This device features a backlit 480x800 resolution screen that supports 4-level grayscale display. Since I just received this development kit, it currently only displays static images of my UI prototype design.
PaperMono, Low-Power, Always-On Dashboard
PaperMono can be used as a low-power, programmable E-Ink information terminal that keeps weather, schedules, and to-dos visible at a glance.
Place it by your bed, on your desk, or beside your equipment. Build your always-on dashboard.
Available August 21.
#M5Stack #PaperMono #EInk #Dashboard #LowPower #IoT #Maker
HanxiaoMeow retweeted
在线体验 lichtspektrum.github.io/lian…
代码 github.com/Lichtspektrum/lia…
使用 Codex 构建
Have some fun with the #M5Stack PaperColor's dithering algorithms. :p
Open-source at: github.com/MarsTechHAN/Paper…
HanxiaoMeow retweeted
One of the cool swag items I received from the Code With Claude Conference was a Cardputer along with a link to a builders challenge page. I hadn’t heard of these tiny devices before, but they seem like fun little microcontrollers for hobby projects.
HanxiaoMeow retweeted
给大家说下目前使用 DeepSeek-V4 (pro/flash) 的最需要注意的问题. 本身其实并不算 bug, 但是却很致命.
问题大概是这样的, 在请求 DeepSeek API 或者 terminal coding agent (claude code, kimi cli 等) / AI IDE (cursor 等) 用 DeepSeek 的时候偶尔会遇到报错:
HTTP 400
{"error":{"message":"The `reasoning_content` in the thinking mode must be passed back to the API.","type":"invalid_request_error","param":null,"code":"invalid_request_error"}}
这个报错的意思是, 请求 DeepSeek API 必须在 tool_call 的时候回传 reasoning_content 这个字段. 听上去没问题, 开了思考模式那肯定要把 reasoning_content 作为上下文回传.
但是来了, 如果任务的这一步制定的 tool_call 过于显而易见, deepseek 返回的 reasoning_content 其实是空字符串. 这就导致了有些写代码的 IDE 直接过滤掉了这个字段, 不回传, 导致 DeepSeek API 报错, 编码任务或者 Agent 就直接挂了.
DeepSeek-V4 API会不会真的有的时候 reasoning_content 空字符串? 答案是会的, 我专门构建了个 POV 场景, 复现概率高达 59%.
那么出现 reasoning_content 为空字符串的时候该怎么办?
经过验证, 答案是必须原样传回去. 即也在 context 中保留这个值为空字符串的字段. 不能是空对象, 也不能丢掉.
那就原样传回去呗? 废什么话呀?
关键是, 现有的各种 terminal coding agent 或者 AI IDE 这并不是默认行为, 它们大部分的默认行为是直接把字段丢掉了, 导致 DeepSeek-V4 API 报错.
所以现在的解决方法是, 要么等 IDE 的官方修复, 要么你用的 IDE 或者 coding agent 是开源的, 自己 fork 一个版本魔改.
另外, 如果你的 Agent 项目要使用 DeepSeek-V4 也要注意这个坑. 避免运行到一半直接报错退出.
以及, 报错重试不太行的, 因为 DeepSeek-V4 在我 POV 这个场景, 59% 的概率都会为空. 如果重试次数为 3, 那偶尔都不够用. 所以还是老实的把问题解决为好.
#deepseek #deepseekv4
HanxiaoMeow retweeted
#BuildInPublic #M5Stack
今天晚上给基于 @M5Stack 的 SickS3 开发的 VibeCoding 工具添加了 2 个状态的动画
1. 处理过程的动画使用的是小猫拉屎💩,小猫拉屎怎么不算是产出呢
2. 休眠状态,30s 不操作进入,是小猫打盹💤的动画
3. 调整了下 Web 对于亮度🔅调解的优化,大部分低端屏幕做不到无极调光,所以就分了几个等级
HanxiaoMeow retweeted
🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length.
🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today!
📄 Tech Report: huggingface.co/deepseek-ai/D…
🤗 Open Weights: huggingface.co/collections/d…
1/n































